Attribution metric analysis Canada US data: close workflow

What can a quarterly match actually establish?

Attribution metric analysis Canada US data calls for this distinction: only the financial result can be confirmed by reconciling billing records; attribution remains limited by the identifiers and other signals observed. Canadian and US reports can therefore differ without an accounting error, while the cause must be tested against each metric's source and definition.

Start with the source records and written metric rules, then test financial reconciliation separately from attribution coverage. Billing or CRM totals can show whether orders and revenue match a defined population. Identifier availability, modeled conversions, and audience units describe what the measurement system observed, so a successful financial test does not settle those questions. Avoid turning either result into a performance claim.

Which records should agree before I look at attribution?

Compare billing and warehouse records in Canadian dollars before checking a US-dollar display. In the Ontario subscription worksheet, billing is recorded in CAD, while the USD figure is a separate calculation rather than the source total for quarterly close.

Retain identifiers, dates, currency, refund records, and the extraction cutoff with the order set. Rebuild the same population in CAD, compare billing with the warehouse, and only then calculate a presentation amount. This sequence helps separate a conversion difference from a mismatch in the underlying records or in the population selected.

Define subscription revenue as charges excluding GST/HST, less refunds entered by the extraction cutoff. GST/HST means Goods and Services Tax/Harmonized Sales Tax. Keep tax lines outside revenue but retain them for payment reconciliation; otherwise a tax-inclusive payment may be compared with a tax-exclusive dashboard measure.

How should I calculate the US-dollar display?

Choose and disclose one consistent conversion convention. One option uses the Bank of Canada daily exchange rate for the billing posting date, or the latest published daily rate on or before that date when the relevant day's rate is unavailable.

The quoted rate is Canadian dollars per US dollar, so divide the CAD amount by that quotation to calculate USD. This is a reporting convention, not a required accounting rule. Record the rate's effective date, and reconcile booked USD ledger amounts separately instead of silently substituting a dashboard conversion for them.

What does each metric source let me verify?

Use business records to check revenue and orders, observed outcomes and coverage to assess modeled conversions, collection records to inspect identifiers, and each source's stated unit to interpret audience reach. A control supports only the differences it can test; estimates and reach counts are not billing evidence.

For every metric row, record its name, source extract, dictionary version, discrepancy explanation, status, and owner. Use reconciled, explained difference, or unresolved as the status; documenting a difference does not make it a match. For audience rows, audience units show whether a direct comparison makes sense. An incompatible unit is an explained difference, not missing Canadian customers.

What should I check when click identifiers are missing?

Inspect the recorded collection state and the exact identifier the report requires. A denied storage state does not mean every event disappeared, and a country value does not show how collection was recorded; the actual field and its availability need separate checks.

For a GA4 setup using Google Consent Mode, inspect analytics_storage and ad_storage, including granted and denied. Preserve the state attached when each event was collected in collection logs or warehouse metadata used for close; do not assume those fields appear automatically in every export.

Group identifier availability by recorded state, including an unknown group when metadata is absent. A country field does not establish collection state, and a technical granted value does not prove legal permission. Save the model version separately for modeled output, since collection state and the report's required identifier answer different questions.

Consent Mode can send cookieless pings when storage is denied, so denied does not establish that all events vanished. Check the report's required identifier field instead of inferring its presence from a flag. Log the field's actual contents and keep that observation separate from the recorded collection state.

What does the Ontario-Michigan case tell me?

It shows that a financial result can match while identifier-based attribution remains incomplete. A US subscription brand sells into Ontario through a .ca domain and joins signups to a warehouse in Oregon.

Ontario revenue matches billing, but paid social reporting lacks click identifiers for visitors who declined tracking. That difference can look like a creative problem, but the appearance alone does not establish a cause. Reconciliation identifies an observation gap while confirming that the financial outcome matches business records.

The evidence does not prove that creative performs equally well in both regions. It shows that the regional attribution comparison covers different observed populations and cannot, by itself, explain a difference as campaign performance. Qualify the comparison rather than drawing a creative conclusion from it.

What details make a metric definition reproducible?

A useful dictionary entry specifies how a count is built, not merely what its label says. Keep revenue and identified clicks as separate entries because their populations and control records differ. Tie each definition to its source, population, treatment, and comparison so another analyst can reproduce the dashboard count.

These field names are illustrative dictionary identifiers, not promised native platform columns. Map them to actual source fields and retain the mapping with the extract. Preserve storage-location and approved-use metadata too. Without a mapping, leave the row unresolved because the next analyst cannot reconstruct what the dashboard counted.

Does a matched result allow a new Canadian data use?

No. This reconciliation assumes a data flow that has already been approved; it does not authorize a new transfer or marketing use. Keep the close within that approved scope, and treat the approval reference as evidence about the existing flow rather than permission to expand it.

The approval reference should identify the Personal Information Protection and Electronic Documents Act, or PIPEDA, where applicable, overseen by the Office of the Privacy Commissioner of Canada, and Quebec Law 25 requirements when Quebec records are included. These notes describe the existing cross-border data flow and do not grant permission to extend its scope.

For the Oregon warehouse, record the existing processor-agreement reference and approved consent or use basis rather than treating a successful join as permission. Assess whether the Oregon Consumer Privacy Act applies to the organization and processing; storage in Oregon alone does not establish that it applies. The Oregon Attorney General enforces the law. Canada's Anti-Spam Legislation, known as CASL, concerns commercial electronic messages; an analytics collection flag is not evidence of consent to send them. Keep the privacy rules compared in view for transfer and permission requirements.

How can another analyst repeat the quarterly close?

Run the process each quarter and review the dictionary whenever a definition changes. Leave an artifact that survives a dashboard refresh and preserves that period's evidence. Campaign measurement practices offer a reference for recording the metric contract before close.

Do not use last quarter's dashboard as the main control. It may repeat an extraction problem or preserve an old definition, so test the current close against source records and the current dictionary. This guards against plausible-looking figures that were generated on different bases.

How should I document a change in the rules?

Keep the earlier definition visible whenever a rule changes instead of overwriting the evidence of a break. The change log should name the affected metric, effective date, old rule, new rule, reason, and treatment of historical comparisons so each figure remains attached to its governing definition.

To test reproduction, have another analyst select the same population from saved extracts, apply the dictionary's tax and refund rules, and recreate the CAD revenue total. Once documented exceptions are resolved, selected order identifiers should match across billing and warehouse records. If they do not, use the attribution mistakes checks before treating the difference as resolved.

The analyst should also recreate the USD display from saved rate dates and rebuild identifier-availability groups. If either calculation fails, leave the worksheet unresolved instead of absorbing the gap into an undocumented adjustment. The saved dates and groupings provide the inputs needed to repeat both checks and identify where reproduction stopped.

The close note should state what matched, which differences were explained, and what remains open. Matching Ontario revenue confirms financial reconciliation, not channel credit, advertising causation, or the quality of the Michigan comparison. Keep that distinction beside the result so it is not mistaken for a performance conclusion.

What should I do first?

Start with the source records and definition of the quarterly population. Reconcile the financial result in CAD before converting it, then assess attribution coverage using recorded collection state and the identifier required by the report. Qualify unresolved comparisons in the close note.